Let a pinned minute read runs from the whole day

/observability/runs gains since/until, so the throughput chart's pin asks
the server for its minute instead of filtering a fixed recent list. This
engine writes ~60 runs a minute, so any minute but the newest read empty.

since is inclusive and until exclusive, matching the minute buckets the
charts are drawn from. Hover stays the client-side preview it was:
scrubbing a day would otherwise be a request per minute rested on.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017MeiWk3Yq12n2pTvnQWYvt

NOTEPAD.md

@@ -51,7 +51,6 @@ is what M4 still waits on, together with porting the flows.
 - CHORE/API: the metrics collector is a bus subscriber, so a storm that overflows the bus queue undercounts. The events dropped are the same ones the websocket drops; exact accounting would need the collector to be fed from the engine rather than the bus.
 - CHORE/API: `/observability/summary` reports the work queue's `depth` as the Redis stream length, which is the journal size (capped at `STREAM_MAXLEN`) rather than a backlog. The health screen shows `pending` instead; the field name still invites the wrong reading.
 - FEAT/UI: the health screen's window is fixed at 24 hours and the charts fold minute buckets in Python. A range picker (and `date_bin()` behind it) is the next step if anyone wants a week.
-- FEAT/API: `/observability/runs` takes no time range and caps at 200 rows, so picking a minute on the throughput chart can only filter the runs the list happens to hold — on a busy engine that is its last minute or so, and any earlier minute reads as empty. A `since`/`until` parameter would let a moment on a chart reach the whole day. Failures are sparse enough that the same pairing works there.
 - CHORE/FLOW: run records for a deleted flow stay until the retention window passes, so a flow that no longer exists keeps appearing in the history. Deliberate — it is a record of what ran — but `forget_flow` could offer to clear it.
 - CHORE/API: nothing can ask the collector to flush now, so anything needing the tables to be current has to wait out `FLUSH_INTERVAL_S` — which is what the soak harness does before clearing its own rows.
 - BUG/UI: `MemoryWorkQueue.stats()` hard-codes `pending: 0`, so the health tile always reads zero on a stack without Redis.

backend/app/api/routes/observability.py

@@ -97,6 +97,11 @@ def _since(hours: int) -> datetime:
     return datetime.now(timezone.utc) - timedelta(hours=hours)
 
 
+def _aware(when: datetime) -> datetime:
+    """A bound as the columns store it. A naive one is read as UTC."""
+    return when if when.tzinfo else when.replace(tzinfo=timezone.utc)
+
+
 @router.get("/summary", response_model=HealthSummary)
 async def read_summary(
     request: Request, controller: FlowControllerDep, session: SessionDep
@@ -275,14 +280,24 @@ def read_runs(
     session: SessionDep,
     flow: str | None = None,
     status: str | None = None,
+    since: datetime | None = None,
+    until: datetime | None = None,
     limit: int = 50,
 ) -> Any:
-    """Recent cascades, newest first."""
+    """Recent cascades, newest first.
+
+    ``since`` is inclusive and ``until`` exclusive, so a window of one minute
+    holds exactly the runs of the minute bucket the charts are drawn from.
+    """
     statement = select(FlowRun).order_by(col(FlowRun.started_at).desc())
     if flow:
         statement = statement.where(col(FlowRun.flow) == flow)
     if status:
         statement = statement.where(col(FlowRun.status) == status)
+    if since:
+        statement = statement.where(col(FlowRun.started_at) >= _aware(since))
+    if until:
+        statement = statement.where(col(FlowRun.started_at) < _aware(until))
     return list(session.exec(statement.limit(min(limit, 200))))
 
 

backend/tests/api/routes/test_observability.py

@@ -102,3 +102,35 @@ def test_the_history_reads_back(
     dead = client.get(f"{PREFIX}/dead-letter", headers=superuser_token_headers)
     assert dead.status_code == 200
     assert isinstance(dead.json(), list)
+
+
+def test_runs_narrow_to_one_minute(
+    client: TestClient, superuser_token_headers: dict[str, str], db: Session
+) -> None:
+    """A minute picked off a chart reaches past what the recent list holds."""
+    minute = datetime.now(timezone.utc).replace(second=0, microsecond=0) - timedelta(
+        hours=3
+    )
+    db.add(FlowRun(id="minute-in", flow=FLOW, started_at=minute, status="ok"))
+    db.add(
+        FlowRun(
+            id="minute-after",
+            flow=FLOW,
+            started_at=minute + timedelta(minutes=1),
+            status="ok",
+        )
+    )
+    db.commit()
+
+    runs = client.get(
+        f"{PREFIX}/runs",
+        headers=superuser_token_headers,
+        params={
+            "since": minute.isoformat(),
+            "until": (minute + timedelta(minutes=1)).isoformat(),
+        },
+    ).json()
+
+    # The upper bound is exclusive, so the run starting the next minute is not
+    # in this one.
+    assert [run["id"] for run in runs] == ["minute-in"]

frontend/src/client/sdk.gen.ts

@@ -1147,9 +1147,14 @@ export class ObservabilityService {
     /**
      * Read Runs
      * Recent cascades, newest first.
+     *
+     * ``since`` is inclusive and ``until`` exclusive, so a window of one minute
+     * holds exactly the runs of the minute bucket the charts are drawn from.
      * @param data The data for the request.
      * @param data.flow
      * @param data.status
+     * @param data.since
+     * @param data.until
      * @param data.limit
      * @returns RunRow Successful Response
      * @throws ApiError
@@ -1161,6 +1166,8 @@ export class ObservabilityService {
             query: {
                 flow: data.flow,
                 status: data.status,
+                since: data.since,
+                until: data.until,
                 limit: data.limit
             },
             errors: {

frontend/src/client/types.gen.ts

@@ -1072,7 +1072,9 @@ export type ObservabilityReadFlowRollupsResponse = (Array<FlowRollup>);
 export type ObservabilityReadRunsData = {
     flow?: (string | null);
     limit?: number;
+    since?: (string | null);
     status?: (string | null);
+    until?: (string | null);
 };
 
 export type ObservabilityReadRunsResponse = (Array<RunRow>);

frontend/src/components/Health/HealthActivity.tsx

@@ -17,6 +17,7 @@ import {
   failuresQueryOptions,
   healthKeys,
   minuteOf,
+  minuteRunsQueryOptions,
   runsQueryOptions,
   timeseriesQueryOptions,
 } from "./queries"
@@ -174,6 +175,7 @@ export function HealthActivity() {
 
   const { data: series } = useQuery(timeseriesQueryOptions())
   const { data: runs } = useQuery(runsQueryOptions())
+  const { data: pinnedRuns } = useQuery(minuteRunsQueryOptions(runsAt.pinned))
   const { data: failures } = useQuery(failuresQueryOptions())
   const { data: audit } = useQuery(auditQueryOptions())
   const { data: dead } = useQuery(deadLetterQueryOptions())
@@ -201,8 +203,14 @@ export function HealthActivity() {
   ): HistoryPoint[] =>
     points.map((point) => ({ ts: point.ts, value: pick(point) }))
 
+  // A pin is read back from the server, so it reaches a minute the recent list
+  // is nowhere near deep enough to hold. A hover stays the client-side preview
+  // it is: scrubbing a day's chart would otherwise be a request per minute the
+  // pointer rests on.
   const shownRuns =
-    runsAt.at === null
+    runsAt.pinned !== null
+      ? (pinnedRuns ?? [])
+      : runsAt.at === null
         ? (runs ?? []).slice(0, RUNS_SHOWN)
         : (runs ?? []).filter((run) => minuteOf(run.started_at) === runsAt.at)
   const shownFailures =
@@ -282,6 +290,8 @@ export function HealthActivity() {
               <p className="text-sm text-muted-foreground">
                 {runsAt.at === null
                   ? "No runs recorded yet."
+                  : runsAt.pinned !== null
+                    ? "Nothing ran in this minute."
                     : "No run from this minute is in the recent list."}
               </p>
             )}

frontend/src/components/Health/queries.ts

@@ -9,13 +9,10 @@ const REFRESH = 30_000
 /**
  * How deep the run and failure lists are read.
  *
- * Deeper than they are shown: picking a minute off a chart filters these rows
+ * Deeper than they are shown: hovering a minute on a chart previews these rows
  * client-side, and a list holding only the newest handful would have nothing
- * to find for any minute but the current one.
- *
- * ponytail: the runs endpoint caps at 200 and takes no time range, so a busy
- * engine still only covers its last minute or so. A `since` parameter is what
- * would let a moment on the chart reach the whole day.
+ * to show for any minute but the current one. Pinning asks the server for the
+ * minute instead, which is what reaches past this depth.
  */
 const RUN_DEPTH = 200
 const EVENT_DEPTH = 100
@@ -53,6 +50,26 @@ export const runsQueryOptions = () => ({
   refetchInterval: REFRESH,
 })
 
+/**
+ * The runs of one minute, wherever it sits in the day.
+ *
+ * A busy engine writes more runs per minute than the recent list is deep, so a
+ * pinned moment is asked for rather than filtered out of what is already held.
+ * `at` is a minute start, and the window is that minute.
+ */
+export const minuteRunsQueryOptions = (at: number | null) => ({
+  queryKey: ["observability", "runs", at] as const,
+  queryFn: () =>
+    ObservabilityService.readRuns({
+      since: new Date((at ?? 0) * 1000).toISOString(),
+      until: new Date(((at ?? 0) + 60) * 1000).toISOString(),
+      limit: RUN_DEPTH,
+    }),
+  // A minute that has passed does not change, and the current one is refreshed
+  // by the unpinned list anyway.
+  enabled: at !== null,
+})
+
 export const failuresQueryOptions = () => ({
   queryKey: [...healthKeys.events, "failure"] as const,
   queryFn: () =>